AI Music on Spotify: Rules, Royalties, and How to Release

Yes, AI music is on Spotify, and you can put yours there too. Spotify does not ban tracks for being made with AI; its September 2025 policy update says the opposite, that it supports artists using AI creatively while going after misuse. What actually stops most people is the distributor in the middle: DistroKid, TuneCore, CD Baby and Amuse each set their own AI rules, and at the time of writing one of them rejects AI-generated music outright.

• Spotify allows AI-assisted and AI-generated music. It removes behaviour, not tools: impersonation, spam uploads and artificial streaming.
• Your distributor is the real gatekeeper. At the time of writing DistroKid and Amuse accept AI music with disclosure, TuneCore applies conditions, and CD Baby does not accept it.
• In the United States, purely AI-generated audio is not copyrightable. The U.S. Copyright Office requires human authorship.
• Spotify pays on streamshare, not a fixed rate per stream, so "how much per stream" has no single answer.
• Disclose AI use where the upload form asks, never clone a real artist's voice, and upload one good song instead of fifty filler ones.

Can you upload AI music to Spotify?

You can, with two conditions: you need the rights to everything in the recording, and you need a distributor willing to deliver it. Spotify does not accept uploads directly from artists. Every track on the service arrives through a distributor or label, which means Spotify's rules and your distributor's rules both apply, and the distributor's are usually stricter.

Spotify's own position has been consistent. In its September 25, 2025 announcement, "Spotify Strengthens AI Protections for Artists, Songwriters and Producers," the company said the industry needs a nuanced approach to AI transparency rather than being forced to classify every song as either "is AI" or "not AI." Music made with AI stays eligible for streaming, playlisting and royalties. What gets pulled is a category of behaviour, and that distinction matters more than any label on your release.

So the honest answer to "can you upload AI music to Spotify" is yes, if you do it like a musician rather than like a content farm. A single well-made track with real artwork, a real artist name and accurate credits moves through the pipeline without friction. A hundred near-identical 31-second uploads under a name that resembles a famous act will not.

What Spotify's AI policy actually says

Spotify's AI policy has three parts, and it's worth understanding each one because they fail in different ways.

Impersonation. Unauthorised AI voice clones, deepfakes and vocal replicas are not allowed. Spotify's stated rule is that vocal impersonation is only permitted when the impersonated artist has authorised it. The platform also invested in stopping fraudulent uploads to other artists' profiles, where someone delivers a track to a real artist's page and collects the streams. This is the hardest line in the policy and there is no grey area on it.

Spam. Spotify launched a music spam filter targeting mass uploads, duplicates and near-duplicates with altered metadata, SEO manipulation inside track and artist metadata, and artificially short tracks designed to cross the royalty-bearing threshold. Flagged uploaders stop being recommended across Spotify programming. In the same announcement Spotify said it had removed more than 75 million spammy tracks over the previous twelve months.

Disclosure. Spotify backed an industry AI disclosure standard through DDEX, the organisation that maintains the metadata plumbing the music business already runs on. The disclosure is granular: it can say AI was used for vocals, for instrumentation, or for post-production, rather than stamping a single binary flag on the release. Spotify began rolling out AI Credits in beta on April 16, 2026, starting with DistroKid deliveries, with the disclosure appearing in a track's Song Credits.

A fourth piece arrived later. On August 11, 2026 Spotify introduced an "AI Persona" badge for artist profiles that represent an AI-generated identity rather than a real person; artists can self-disclose in Spotify for Artists, and Spotify's own review team also applies the badge to profiles it identifies. By default, Spotify says it will not include AI Personas in editorial or algorithmic recommendations unless a listener follows them. That is a label on the artist identity, not on an individual track made with AI tools.

Two things follow from this. First, track-level disclosure is currently self-reported, so it depends on creators being honest. Second, and more usefully for you: disclosing does not get your song demoted. Spotify has said it does not penalise or down-rank music for being AI-assisted. Treat the AI disclosure box the way you treat the songwriter credit field, as metadata you fill in accurately.

Which distributors accept AI music?

A distributor is the service that delivers your finished audio and metadata to Spotify, Apple Music and the rest, and collects the royalties on your behalf. This is where AI releases actually get stopped. Policies here changed several times during 2025 and 2026, so treat the table below as a snapshot at the time of writing and read the current terms before you pay for a release.

Distributor	AI music accepted?	What it asks of you
DistroKid	Yes, with disclosure	Its help centre says you must own 100% of the rights, including the legal right to distribute music created with any AI tools, samples or lyrics, with no impersonation and no mass-generated spam. AI Credits disclosure is part of the upload flow: lyrics, music, all of the audio, or part of the audio.
TuneCore	Conditionally	Its GenAI Music Content Framework says it only distributes music made with generative AI models trained on fully licensed datasets; music from models that rely on unlicensed datasets is not eligible. That test applies if generative AI was used at any point in the process, not only for fully generated tracks.
CD Baby	No	CD Baby's Help Center states it cannot accept any AI-generated content, even if it is commercially licensed, and even where you contributed original sounds and only part of the recording is AI. It says it will remove such content if it finds it. The separate Downtown Music content policy also prohibits AI content trained on unlicensed recordings and AI content that mimics popular artists.
Amuse	Yes, with exclusions	Its support pages say AI-generated music is allowed, but Qobuz, Meta and YouTube Content ID exclude such releases. Tracks that imitate or replicate well-known artists, and AI covers, are rejected.

The pattern is clear enough. Distributors are not asking whether a machine was involved; they're asking whether you can prove you have the right to sell the result. That means keeping your paper trail: the plan you were on when you generated the track, the date, the export, and the tool's licence terms. If a compliance check lands six months after release, that folder is what resolves it in a day instead of a month.

How to release a song on Spotify, step by step

People search "how to make a song on Spotify" meaning two different things: writing and recording the song, and getting a finished file onto the platform. Here is the second half, which is the part with rules.

1. Finish the song properly. Export the highest quality file you have, ideally WAV at 16-bit/44.1 kHz or better. Distributors accept MP3, but you only get one shot at the master that sits on the platform for years.
2. Pick your artist name and check it is free. Search Spotify and the trademark register. A name that collides with an existing act causes profile-matching problems later, and a name designed to be mistaken for a famous artist is an impersonation risk.
3. Make artwork you own. Square, 3000x3000 pixels — the size Spotify recommends, though it accepts down to 640x640 — no stock photos you haven't licensed, no logos, no text that misstates the release. Artwork is one of the most common rejection reasons and one of the easiest to get right.
4. Choose a distributor from the table above, based on whether it accepts your kind of release and whether it charges per release or per year.
5. Upload and fill in the metadata carefully. Exact track title, correct featured-artist fields, songwriter and producer credits, language, explicit flag, and genre. Do not pad the title with keywords; that is exactly what the spam filter looks for.
6. Complete the AI disclosure. If the upload form asks whether AI was used for lyrics, vocals or instrumentation, answer honestly. On DistroKid this feeds the AI Credits that appear in Spotify's Song Credits.
7. Set a release date at least three to four weeks out. This gives you time to claim the release in Spotify for Artists and pitch it to editorial playlists. Spotify asks for pitches at least seven days before the release date; miss that window and you also lose your pick for Release Radar.
8. Claim your Spotify for Artists profile and add a photo, a bio and your socials. An unclaimed profile is a dead end for anyone who likes the song.
9. After release, watch the numbers for anything you didn't cause. Sudden spikes from countries you have no audience in are the signature of artificial streaming, and it gets your release pulled whether or not you bought it.

If you are still at the writing stage, our guides on how to write a song and song structure cover the part that happens before any of this.

Can you copyright AI music?

Copyright is the exclusive legal right to control how a creative work is copied, distributed and performed. In the United States, whether AI music qualifies comes down to one requirement: human authorship.

In January 2025 the U.S. Copyright Office published Part 2 of its report Copyright and Artificial Intelligence, on copyrightability. Its conclusions are specific and worth knowing verbatim in substance. Human authorship is a bedrock requirement, so a work generated entirely by AI is not copyrightable. Using AI as a tool to assist human creativity does not remove protection from the resulting work. And critically for anyone using a prompt box: the Office concluded that selecting prompts alone, even detailed ones that took real effort, does not by itself produce a copyrightable work, though it noted this could change as the technology does. Where a human contributes perceptible expression, such as their own lyrics, or a creative selection and arrangement of AI-generated material, that human contribution can be protected. It is assessed case by case.

Two practical consequences. First, "I can't register a copyright on this" is not the same as "I can't release or sell this" — distribution and monetisation are governed by the platform's terms and the AI tool's licence, not by registration. Second, the more of yourself you put into a track — your lyrics, your structural edits, your arrangement of sections — the more of it is yours in the legal sense. If you want to check whether an existing song is protected before you build on it, we walk through the registries in how to check if a song is copyrighted. None of this is legal advice; for a release with money attached, talk to a music lawyer.

How do you make money from music on Spotify?

Streams turn into royalties, the royalties go to your distributor, and the distributor pays you. Several of the large distributors keep 0% of your recording royalties in exchange for a flat fee, while others take a commission, so read your agreement rather than assuming.

There is no fixed per-stream rate, and this is the single most misunderstood thing about streaming income. Spotify's Loud & Clear site explains that payouts work on streamshare: your streams in a market divided by total streams in that market gives your share of that market's royalty pool for the month. That means the same 10,000 streams are worth different amounts depending on where your listeners are and what month it is. Its 2026 edition, covering 2025, reports that artists whose music represented 1 in every 1 million streams earned over $11,000 on average that year, and that Spotify paid the music industry more than $11 billion across 2025. Note what that figure is and is not: it describes a share of all streaming on the service, not a payout for one million plays. The averages also hide enormous variation, so do not budget against them.

Two things follow. Streaming royalties are a long game that rewards catalogue and repeat listeners, not a launch-week payday. And for most independent creators, music on Spotify is one income stream beside sync placements, direct sales, live work and audience-building — not the whole business. If your goal is licensing music for video, AI background music for videos covers a market where a single placement can outearn a year of streams.

What gets AI songs taken down?

Almost every takedown traces to one of four things, and none of them is "you used AI."

• Artificial streaming. Buying streams, bot farms, or playlist schemes that promise guaranteed numbers. Spotify detects it, removes the tracks, and distributors pass on penalties. Third-party promo services that guarantee stream counts are the most common way honest artists get caught in this.
• Impersonation. Cloned vocals of a real artist without authorisation, artist names engineered to be confused with a famous act, or uploads delivered to someone else's profile.
• Spam patterns. Mass uploads, duplicate and near-duplicate versions with shuffled metadata, keyword-stuffed titles, and batches of tracks barely over the length that starts earning royalties.
• Rights problems in the audio. Uncleared samples, a melody or recording substantially similar to a protected work, or lyrics you don't have the right to use. AI output is not exempt from ordinary copyright law.

If you want to know how likely your track is to be spotted as machine-made in the first place, the tells and the detection tools are covered in how to tell if a song is AI-generated.

Best practices for releasing AI music on Spotify

• Own your artwork. Commission it, shoot it, or generate it with a tool whose licence covers commercial use. Reused stock and scraped images are the fastest route to a rejected release.
• Never impersonate. No cloned voices, no "in the style of [famous artist]" in your metadata, no artist name one letter away from someone else's.
• Disclose where the form asks. It costs you nothing in reach and it protects you if a compliance check comes later.
• Add your own creative input. Write the lyrics yourself, rework sections, edit the arrangement, re-sing a part, mix it properly. This improves the song, strengthens your copyright position, and makes the release defensible.
• Release less, better. One track a month with real artwork and a real pitch beats fifty uploads. Volume is the signal the spam filter is built to catch.
• Keep your receipts. Plan, date, tool, licence terms, exported files. Put them in one folder per release.

Where Neume fits

Neume is an AI song generator: you type what you want into a single prompt box — a description, a style, your own lyrics, or lyrics plus a style — and it returns a full song with vocals and instruments, mixed, in about a minute. You can separate the result into vocal and instrumental stems, rewrite the lyrics in specific sections with Remix, and download the track as a 320 kbps MP3 to keep. Commercial use is included, and you own what you generate, which is the licence position your distributor will ask about. There is one free credit when you sign up, so your first song costs nothing.

What Neume cannot do is promise anything about a platform it doesn't control. Nobody can honestly guarantee that a given track will never be flagged, never receive a Content ID claim, or never be removed, and streaming policies change from quarter to quarter — the DDEX disclosure standard and Spotify's AI Credits both arrived inside twelve months. What you get is clear ownership of your side of the deal, files you keep, and the ability to add your own lyrics and edits so there is real human authorship in the release. More on the ownership side in why Neume lets you download and keep your songs, and if you're comparing tools first, Suno alternatives lays out the field.

Release one song properly and the rest gets easier

The gap between AI music that quietly lives on Spotify and AI music that gets pulled is not technical. It's a distributor that accepts your release, an honest disclosure, artwork you own, a name that isn't borrowed, and no shortcuts on streams. Do that once and you have a template you can run every month. Start with the song: write the lyrics yourself, generate a version with the AI song generator to hear how they sit, then edit until it's actually yours.